A Shelter for the Heart: The Impact of Teacher Quality on Students' Mental Health

ABSTRACT

Adolescent mental health has become a growing public health concern. This study examines the impact of teacher quality, proxied by teacher experience, on student mental health. To address potential endogeneity concerns and identify causal effects, we exploit data from a sample of schools where students and teachers are plausibly randomly assigned to classes. Our results indicate that students assigned to head teachers with greater experience exhibit significantly improved mental health. By contrast, we find no such effects for subject teachers who are solely responsible for academic instruction. The mechanism analysis suggests that more experienced head teachers strengthen teacher–student relationships, engage more with parents, report higher work satisfaction, and promote more supportive peer interactions. Our results highlight the important role of school environments in determining adolescents' mental health and offer actionable insights for interventions.

1. Introduction

Mental health has emerged as a growing public health issue, with adolescents particularly at risk. Globally, approximately 10%–20% of adolescents experience mental health disorders (WHO ²⁰²⁰), contributing to 16% of the total disease and injury burden in this age group (WHO ²⁰²¹). Untreated mental health conditions during adolescence frequently persist into adulthood (Johnson et al. 2018; Patel et al. ²⁰¹⁸), impairing academic achievement and resulting in poorer socioeconomic outcomes, such as unemployment and reduced earnings (McLeod and Kaiser ²⁰⁰⁴; Case et al. ²⁰⁰⁵; Currie and Stabile ²⁰⁰⁷; Fletcher ²⁰¹³; Mousteri et al. ²⁰¹⁹). The substantial individual and societal costs associated with adolescent mental health problems underscore the importance of identifying their determinants. Given that adolescents spend a substantial portion of their time at school, the school environment may play a critical role in shaping their mental health. This study examines whether teacher quality, as a key aspect of the school environment, affects adolescents' mental health outcomes.

Teachers are among the most crucial inputs in education (Hanushek 1986; Chetty et al. ^{2014a, 2014b}). While a growing body of literature examines how teacher quality influences students' academic outcomes (Carrell and West 2010; Staiger and Rockoff ²⁰¹⁰; Chetty et al. ^{2014a, 2014b}; Gerritsen et al. ²⁰¹⁷; Pereda‐Fernández ²⁰¹⁷; Hanushek et al. ²⁰¹⁹; Bietenbeck et al. ²⁰²³), evidence regarding the impact of teacher quality on student mental health remains limited. Existing research indicates that higher‐quality teachers improve students' human capital outcomes by fostering stronger teacher–student relationships and promoting effective teaching practices (Araujo et al. 2016; Thijssen et al. ²⁰²²). These mechanisms may plausibly extend to students' mental health. This study contributes to the literature by examining whether teacher quality improves students' mental health.

Using data from the China Education Panel Survey (CEPS), we investigate the impact of head teacher quality on junior high school students' mental health. ¹ Specifically, we estimate the effect of teacher experience, a common observable dimension of teacher quality and productivity (Papay and Kraft 2015; Gerritsen et al. ²⁰¹⁷), on student mental health outcomes in China. ² Studying this relationship faces two empirical challenges. First, teachers and students are typically matched non‐randomly. Schools may assign more experienced or higher‐quality head teachers to classes with more severe mental health problems, generating sorting between teacher and student characteristics. Second, parental selection may arise if parents choose schools or classrooms based on perceived teacher quality, potentially favoring teachers with more experience. To address these concerns, we analyze a sample of schools that plausibly randomly assign students to classrooms, ensuring plausibly random student–teacher pairings (Gong et al. 2018). By controlling for school fixed effects, our identification strategy compares mental health outcomes across students assigned to head teachers with varying levels of experience within the same school. This random assignment allows us to identify the causal effect of head teacher experience on student mental health.

Our findings indicate that head teacher experience significantly improves student mental health. Specifically, each additional year of head teacher experience increases students' average mental health index and the principal component analysis (PCA) index by approximately 0.03 standard deviations. We find no evidence of a concave or inverted‐U relationship between experience and effectiveness. Estimates based on experience categories indicate that the effects persist and generally increase with experience, consistent with cumulative returns. Using follow‐up data collected 1 year later, we continue to find positive effects on mental health, suggesting that the gains persist into the medium term. Our heterogeneity analysis highlights that these benefits are particularly pronounced for students from low socioeconomic status families and those in rural schools, suggesting that experienced head teachers play a crucial role in supporting the mental health of students from disadvantaged socioeconomic backgrounds. To distinguish head teachers from subject teachers, we test whether the experience of subject teachers who do not serve as head teachers affects student mental health. We find no statistically significant effects, underscoring the distinctive pastoral and managerial roles of head teachers.

The mechanism analysis suggests that the positive effect of head teacher experience on student mental health operates through multiple channels, including stronger teacher–student relationships, more frequent teacher–parent interactions, higher work satisfaction, and more supportive peer interactions. To quantify the relative importance of these pathways, we apply the decomposition method developed by Gelbach (2016) and Gong et al. (2018). The results indicate that these mechanisms collectively explain approximately 70% of the total effect of head teacher experience on student mental health. We also find suggestive evidence that experienced head teachers provide more supportive responses to students' mental health problems.

Our study contributes to the literature in three ways. First, it adds to research on the determinants of adolescent mental health. Previous studies have identified several factors influencing adolescent mental health, including educational attainment (Weng 2025), body weight (Revah‐Levy et al. ²⁰¹¹), parental health (Le and Nguyen 2017), family income (Baird et al. 2013; Akee et al. ²⁰¹⁸), early‐life conditions (Persson and Rossin‐Slater ²⁰¹⁸), religion (Fruehwirth et al. 2019), neighborhood environment (Kling et al. 2007), and social media (Fruehwirth et al. 2024). Closely related to our work are studies that examine how the school environment and peer composition influence student mental health (Song 2017; Kiessling and Norris ²⁰²²; Giulietti et al. ²⁰²²; Linder et al. ²⁰²⁵; Chen et al. ²⁰²⁶). We extend this work by exploring whether teacher quality, a central aspect of the school environment, also affects students' mental health. We find that greater teacher experience significantly improves student mental health, with effects that persist over the medium term.

Second, our study contributes to the literature examining the effects of teacher quality on students' human capital outcomes, where teacher quality is proxied using cognitive skills (Hanushek et al. 2019), teacher experience (Gerritsen et al. 2017; Huang et al. ²⁰²³), subject knowledge and didactic skills (Bietenbeck et al. 2023), teacher credentials and educational level (Clotfelter et al. 2010; Qin et al. ²⁰²⁵), student achievements (Chetty et al. ^{2014a, 2014b}), or a combination of multiple measures (Carrell and West 2010; Staiger and Rockoff ²⁰¹⁰). Using a nationally representative dataset with random teacher–student assignment in China, our study provides novel evidence on the relationship between teacher quality and student mental health. ³ Our focus on teacher quality and student mental health is closely related to Qin et al. (2025), who show that teachers' educational attainment improves students' mental health. We provide complementary evidence by linking teacher quality to student mental health through a distinct dimension, teacher experience. While education captures a largely pre‐service or early‐career stock of formal human capital, experience reflects learning‐by‐doing and the accumulation of role‐specific skills, such as classroom management, emotional support, and relationship building, that may not be fully reflected in credentials. ⁴ Importantly, teacher experience, our primary explanatory variable, is not included in the baseline specifications of Qin et al. (2025). These dimensions correspond to different policy levers: credential requirements and training pipelines versus teacher retention, assignment, mentoring, and leadership‐role allocation. Conceptually, teacher experience may affect student mental health through dynamic relational competencies that develop with repeated interactions, such as teacher–student trust formation, parent communication routines, and conflict resolution, rather than primarily through subject knowledge or pedagogical content knowledge associated with formal education. Consistent with this view, we examine mechanisms related to the head teacher's role, including teacher–student relationships, teacher–parent interactions, teacher job satisfaction, and peer interactions, thereby extending the mechanism evidence in Qin et al. (2025). While Qin et al. (2025) use follow‐up data to estimate the medium‐term effects of Grade 7 teachers' education on Grade 8 mental health, we estimate the short‐run effect of Grade 7 head‐teacher experience on Grade 7 mental health and further show that these effects remain statistically significant in Grade 8. Finally, we apply the method proposed by Oster (2019) to evaluate the robustness of our main results to potential bias from unobserved teacher characteristics that may correlate with teacher experience.

Third, our study contributes to the literature on the effects of teacher experience on student human capital outcomes. Existing research on teacher experience has primarily focused on academic performance and yielded mixed results. While some studies find positive effects (Papay and Kraft 2015; Gerritsen et al. ²⁰¹⁷; Huang et al. ²⁰²³), others report no significant impact (Rivkin et al. 2005; Cabrera and Webbink ²⁰²⁰). We extend this literature by examining mental health as another critical dimension of human capital outcomes. Our results indicate that teacher experience significantly enhances student mental health, suggesting that the benefits of teacher experience extend beyond academic achievements. Furthermore, our finding that subject teacher experience has no significant effect on student mental health highlights the importance of teacher roles. These results suggest that not all teaching experience equally contributes to students' well‐being, and that the broader pastoral and managerial responsibilities undertaken by head teachers are critical for shaping students' mental health.

The paper is organized as follows. Section 2 describes the institutional background, while Section 3 discusses the data. Section 4 presents the empirical strategy and balance tests. Section 5 reports the main empirical results. Section 6 investigates potential mechanisms. Section 7 provides robustness checks, and Section 8 concludes the paper.

2. Institutional Background

2.1. China's Educational System and Random Class Assignment

In April 1986, China enacted the Compulsory Education Law of the People's Republic of China, which has since undergone multiple revisions. This foundational law dictates that junior high school admissions must be free of entrance examinations and tuition fees; students are instead assigned to schools based on their location of household registration (hukou). A crucial distinction in the Chinese system is the prohibition of academic tracking in junior high schools, a policy explicitly enforced by the Ministry of Education since 2006. ⁵ While student tracking is common in senior high schools, junior high schools aren't allowed to form “elite” or “non‐elite” classes, which helps ensure equitable educational opportunities. Consistent with this objective, the Ministry of Education has promoted random assignment of students to classrooms as a mechanism for achieving an equitable distribution within the 9‐year compulsory system (6 years of primary school and 3 years of junior high school). As a result, students are plausibly assigned to a class upon entry to Grade 7 and typically remain in that class through Grade 9.

In practice, junior high schools in China typically adopt one of two random assignment strategies: purely random assignment and balanced random assignment. Purely random assignment uses computer algorithms to randomly place students based on demographic and classroom characteristics. By contrast, balanced random assignment keeps average academic performance, typically proxied by prior test scores, similar across classrooms while maintaining randomness in other dimensions. Although our data do not allow us to distinguish between these two approaches, both are widely recognized as valid strategies for identifying causal teacher effects. Importantly, institutional incentives strongly encourage adherence to these assignment protocols. Schools that claim to implement random assignment face reputational risks if found to deviate, including potential parental complaints to local education bureaus or negative media attention.

This external pressure for transparency is matched by internal mechanisms that prevent teachers from undermining the randomization process. While teachers' performance evaluations, career advancement, and bonuses are often tied to student test scores, which creates individual incentives to prefer higher‐ability cohorts, several institutional features prevent teachers from influencing assignment and maintain randomization in practice. First, assignment decisions are centralized within the school administration rather than determined by individual teachers, thereby limiting the capacity for teachers to manipulate student placements. Second, if assignment were negotiable, competing teacher demands would create internal conflicts. Random assignment provides a politically neutral solution that ensures institutional stability and is widely accepted as equitable by the faculty. Third, any discretionary deviation that systematically advantages some teachers would be perceived as inequitable by their colleagues, likely generating complaints and eroding workplace morale. To maintain internal equity, avoid faculty disputes, and preserve the legitimacy of the assignment process, school administrators have strong incentives to adhere to transparent and uniform random assignment rules that apply equally to all teachers.

2.2. The Role of Head Teachers and Student Mental Health

The head teacher system is mandated nationwide by the Ministry of Education under the 2009 Regulations on the Work of Primary and Secondary School Head Teachers and has long been a standardized feature of China's compulsory education system. ⁶ In Chinese junior high schools, each class is assigned a single head teacher who is responsible exclusively for that class and is typically selected from teachers of core subjects (Chinese, mathematics, and English). Assignments are made at the start of each grade level (Grade 7 in junior high school) and generally follow random or stratified‐random procedures within schools to ensure balanced class composition. Head teachers occupy a central institutional role in students' daily lives, serving not only as academic instructors but also as mentors, counselors, and primary intermediaries between the school and the family (Gong et al. 2018; Huang et al. ²⁰²³). National regulations specify their responsibilities, which include monitoring students' academic progress and personal development; managing classroom organization and discipline; organizing class activities and meetings; fostering positive and inclusive peer environments; attending to students' psychological well‐being and providing pastoral support; maintaining regular communication with parents; and coordinating with families and communities to support student development. Unlike subject teachers, whose responsibilities are confined to specific curricular areas, head teachers are institutionally accountable for students' holistic development, encompassing academic achievement, behavioral adjustment, social relationships, and mental health. This pastoral role is formally embedded in nationwide job descriptions, performance evaluations, and professional training requirements.

The institutional structure of junior high schools thus places head teachers at the forefront of adolescent development. This role becomes especially critical in the context of China's intense academic environment, where competitive entrance exams for elite senior high schools and subsequent university access exert considerable pressure on students (Guo and Qu 2022). As a result, adolescent mental health has emerged as a pressing concern. By 2021, the age‐standardized prevalence of mental disorders among Chinese children and adolescents had reached 8.9%, affecting an estimated 30.8 million individuals (Dong et al. 2025). This figure reflects the combined prevalence of 10 disorder categories, including depressive disorders (0.4%), anxiety disorders (2.6%), schizophrenia (0.02%), autism spectrum disorders (0.7%), bipolar disorder (0.05%), conduct disorder (1.3%), eating disorders (0.09%), attention‐deficit/hyperactivity disorder (3.6%), idiopathic developmental intellectual disability (0.4%), and other mental disorders (0.09%). In this high‐stakes educational climate, head teachers are expected to provide not only academic instruction but also emotional support. Their duties include holding regular class meetings, conducting one‐on‐one conversations with students, and maintaining close communication with parents about students' academic performance and mental health. This institutional arrangement offers a compelling setting to investigate how teacher quality, particularly head teacher quality, affects adolescent mental health outcomes.

3. Data

3.1. Data Source

We use data from the CEPS, a nationally representative survey of junior high school students in China, designed and implemented by Renmin University of China. The baseline survey, conducted during the 2013–2014 academic year, employed a multi‐stage probability‐proportional‐to‐size sampling strategy. The survey includes data from 28 county‐level units, 112 schools, and 438 classes, covering approximately 20,000 students in grades 7 and 9. ⁷ Within each selected school, two classes per grade were randomly sampled. The dataset provides rich information on student demographics, as well as detailed characteristics of their parents, teachers, and school administrators.

For this study, we restrict the sample to 7th‐grade students to mitigate potential selection bias stemming from student transfers in later grades, particularly Grade 9. This restriction is necessary because the available data do not contain information on class assignments for 9th‐grade students during their early junior high school years. These exclusions yield an analytic sample of 10,279 7th‐grade students. In the 2013–2014 CEPS baseline wave, Grade 7 students were surveyed in two periods. Approximately 61% of classes were surveyed in autumn 2013 (roughly three months after the start of Grade 7), while 39% of classes were surveyed in spring 2014 (approximately seven months after Grade 7 begins).

Our primary analysis relies on Grade 7 classrooms in the 2013–2014 CEPS baseline wave. We utilize the 2014–2015 follow‐up, which tracks the same cohort into Grade 8, only for the supplementary analyses presented in Section 5.2. Although this follow‐up wave provides more detailed questions on mental health, we restrict our primary analysis to the 2013–2014 baseline wave to maintain the validity of our identification strategy. Student mobility between Grade 7 and Grade 8 introduces endogenous selection that would violate our random assignment assumption. By the 2014–2015 academic year (Grade 8), some students may have transferred schools or changed classes, introducing potential endogenous sample selection. Importantly, head teacher experience may both affect students' mental health and influence their likelihood of remaining in the same class or school. For example, more experienced head teachers may improve mental health and school satisfaction, reducing transfers or requests for class changes. The remaining students in classrooms with more experienced head teachers would then be positively selected, weakening the random assignment implemented at the start of Grade 7. Even if head teachers do not change, classroom composition may become correlated with head teacher experience. Our identification strategy relies on random assignment of students to head teachers within schools. Once students selectively transfer based on teacher characteristics, this randomness is violated. Accordingly, focusing on Grade 7 outcomes immediately after assignment provides a more robust setting for causal inference. This approach captures the effects of the treatment before selective mobility can compromise the plausibly random allocation of students to head teachers.

3.2. Variables

3.2.1. Head Teacher Characteristics

Head teachers play an important role in students' daily lives. While subject teachers are primarily responsible for delivering instruction and ensuring that students meet academic goals, head teachers oversee all aspects of classroom management and are actively involved in students' social development. Consequently, they tend to exert a stronger influence on students' mental health than subject teachers. To measure teaching experience, we use responses from the CEPS teacher questionnaire, which asks head teachers to report the number of years they have worked as teachers. Our analysis also controls for additional head teacher characteristics, including age, gender, years of schooling, professional title, and indicators for whether the head teacher teaches one of the three core subjects (Chinese, Mathematics, and English), holds a tenured position, occupies an administrative position, or has previous experience working at other schools.

3.2.2. Measure of Student Mental Health

The CEPS asks students five questions about their mental health, specifically whether they have experienced the following feelings in the past 7 days: feeling blue, unhappy, perceiving life as meaningless, sad, and depressed. ^{8 , 9} Responses are recorded on a five‐point scale, ranging from 1 (“always”) to 5 (“never”), with higher scores indicating better mental health. The structure and content of these questions closely resemble the Mental Health Inventory‐Five Item (MHI‐5), a clinically validated subscale of the SF‐36 Health Survey (Berwick et al. 1991). ¹⁰ The MHI‐5 has been extensively validated as an effective screening instrument for detecting mental health problems in large populations (Yamazaki et al. 2005). Research has demonstrated that the MHI‐5 performs comparably to, and in some cases outperforms, other widely used self‐reported mental health instruments, including the General Health Questionnaire (McCabe et al. ¹⁹⁹⁶; Hoeymans et al. ²⁰⁰⁴; Kelly et al. ²⁰⁰⁸) and the 28‐item Somatic Symptom Inventory (Berwick et al. 1991). This measure and its variants have been utilized extensively in the medical literature (Rivera‐Riquelme et al. ²⁰¹⁹; LaMontagne et al. ²⁰²¹) and economic research (Stillman et al. 2009; Green ²⁰¹¹; Jung et al. ²⁰²⁵).

In the Chinese educational context specifically, these five CEPS items have been employed in prior studies examining student mental health (Gong et al. 2018; Xu et al. ²⁰²²; Chen et al. ²⁰²⁶). Based on the five items, we construct two indices of mental health. The first is the Average index of mental health, defined as the arithmetic mean of the five items' scores. Following Gong et al. (2018), we standardize this index at the school level to have a mean of zero and a standard deviation of one. The second is the PCA index of mental health, constructed using principal component analysis on the same five items. We retain principal components with eigenvalues greater than one (using the eigenvalue‐one criterion) and similarly standardize this PCA index at the school level. ¹¹

3.3. Estimation Sample

To identify the causal effect of head teacher experience on student mental health, it is essential to account for potential bias arising from non‐random sorting. For example, schools might assign head teachers based on students' mental health, or parents may select classes taught by more experienced teachers. In such cases, the correlation between head teacher experience and unobserved student characteristics could result in omitted variable bias. To address these endogeneity concerns, we employ school fixed effects and restrict the analysis to settings where students are plausibly randomly assigned to classes within schools. Following the approach of Gong et al. (2018) and Megalokonomou and Zhang (2024), we implement a set of criteria based on information from the school principal and teacher survey modules. First, the school principal confirms that students are randomly assigned to classrooms in Grade 7. Second, the principal reports that no student reallocation across classes or grades occurred after the initial assignment. Third, all head and subject teachers within the same grade report that class assignment is not based on students' test scores. Applying these criteria and further limiting the sample to schools with exactly two classes per grade, we identify 43 out of 112 schools as meeting the random assignment conditions, resulting in a sample of 3865 students. After excluding observations with missing data on outcome or control variables, our final analytic sample comprises 3529 students across 86 classrooms in 43 schools.

A potential concern is the external validity of estimates derived from our analytic sample, which is restricted to schools that implement random class assignment. We address this concern using two checks. First, we compare the geographic stratification of our analytic sample with the full Grade 7 CEPS sample. The results in Panel A of Supporting Information S1: Appendix Table A1 show that the distribution of schools across the three primary CEPS strata (national core, Shanghai, and high‐migrant areas) is similar in the two samples. This similarity indicates that our sample preserves the geographic diversity of the national dataset rather than concentrating in specific regions. Second, we compare observable characteristics across four dimensions: school characteristics (Panel A), class and head teacher characteristics (Panel B), student demographics and family background (Panel C), and student mental health (Panel D). The results in Supporting Information S1: Appendix Table A1 show that the two samples are similar across all dimensions. This similarity, combined with matching geographic stratification, indicates that our sample represents the diversity of contexts covered by CEPS, from less developed rural areas to advanced urban centers. While we cannot rule out differences in unobservable characteristics, the close correspondence across geography and measured characteristics strengthens confidence in the external validity of our estimates to the broader Chinese junior high school context.

3.4. Summary Statistics

Table 1 presents the summary statistics. Panel A reports the average and PCA indices of mental health, both standardized to have a mean of zero and a standard deviation of one. Figure 1 displays the density distribution of the two indices, showing considerable variation in student mental health. Panel B of Table 1 summarizes student‐level characteristics. The average student age is 13 years, and approximately 52% are male. Students have 0.6 siblings on average, and 47% hold an agricultural hukou. The average number of grades repeated and skipped in primary school is 0.14 and 0.02, respectively. Approximately 85% of students attended kindergarten. The average years of schooling for fathers are 11, and for mothers are 10.

TABLE 1.

Summary statistics.

	Mean	SD	Observations
	(1)	(2)	(3)
Panel A. Mental health
Average index of mental health	0.000	0.994	3529
PCA index of mental health	−0.000	0.994	3529
Panel B. Student characteristics
Age	13.061	0.886	3529
Male (Yes = 1)	0.519	0.500	3529
Number of siblings	0.634	0.778	3529
Agricultural hukou (Yes = 1)	0.473	0.499	3529
Number of grades repeated in primary school	0.138	0.429	3529
Number of grades skipped in primary school	0.024	0.270	3529
Kindergarten attendance (Yes = 1)	0.846	0.361	3529
Father's years of schooling	10.532	3.227	3529
Mother's years of schooling	9.931	3.410	3529
Panel C. Head teacher characteristics
Experience	13.953	7.945	86
Age	35.930	6.689	86
Male (Yes = 1)	0.279	0.451	86
Years of schooling	15.733	0.873	86
Professional title	2.512	0.979	86
Teaching one of the core academic subjects (Yes = 1)	0.872	0.336	86
Tenured position (Yes = 1)	0.849	0.360	86
Administrative position (Yes = 1)	0.116	0.322	86
Previous experience working at other schools (Yes = 1)	0.547	0.501	86
Teaching a class with more than 50 students (Yes = 1)	0.326	0.471	86

Note: Data are from the 2013–2014 CEPS. Summary statistics in Panels A and B are reported at the student level, while those in Panel C are reported at the class level.

FIGURE 1.

Density distribution of the average and PCA indices of mental health. Data are from the 2013–2014 CEPS.

Panel C of Table 1 provides descriptive statistics for head teacher characteristics. On average, head teachers have 14 years of teaching experience. Figure 2 displays the density distribution of head teacher experience, showing that the majority have between 0 and 20 years of experience, with only a small fraction exceeding this range. The average age of head teachers is 36 years, and 28% are male. Their mean educational attainment corresponds to 16 years of schooling, and the most common professional title is Level‐B Teacher. ¹² Approximately 87% teach one of the core academic subjects, 85% hold a tenured position, 12% hold an administrative position, and 55% have previous experience working at other schools. Moreover, 33% of them teach classes with more than 50 students.

FIGURE 2.

Density distribution of the head teacher experience. Data are from the 2013–2014 CEPS.

4. Empirical Strategy

4.1. Empirical Model

To examine the causal effect of head teacher experience on students' mental health, we estimate the following equation:

$$Y_{ics}=\alpha +\beta {TeacherExp}_{cs}+X_{ics}^{\mathit{\prime }}\mathrm{\Gamma }+Z_{cs}^{\mathit{\prime }}\mathrm{\Pi }+{\phi }_s+{\mu }_{ics}$$ (1)

where Y _ics denotes the average and PCA indices of mental health of student i in classroom c of school s. The variable TeacherExp _cs measures the years of experience of the head teacher assigned to classroom c. Since each class is assigned a single head teacher and students are plausibly randomly allocated to classes within schools, the coefficient β identifies the causal effect of head teacher experience on students' mental health. The vector X _ics represents students' individual and family characteristics, including age, gender, number of siblings, number of grades repeated in primary school, number of grades skipped in primary school, parental years of schooling, and binary indicators for agricultural hukou and kindergarten attendance. The vector Z _cs contains characteristics of the head teacher, including age, age squared, gender, years of schooling, professional title, and indicators for whether the head teacher teaches one of the core academic subjects, holds a tenured position, occupies an administrative position, has previous experience working at other schools, and teaches a class with more than 50 students. ¹³ Given that our identification strategy relies on random class assignment within the same school, we include school fixed effects φ _s . To address potential correlation in student outcomes within schools, we cluster standard errors at the school level.

We examine the extent to which our measure of head teacher experience captures underlying teacher characteristics. Specifically, we assess whether observable teacher characteristics are correlated with experience in ways that could confound our estimates. Supporting Information S1: Appendix Table A2 presents the results from regressions of head teacher experience on various observable teacher characteristics. ¹⁴ The results indicate that teacher age is strongly associated with experience. Although the coefficients on other characteristics are not statistically significant, several are sizable in magnitude. These results highlight the importance of controlling for teacher characteristics when estimating the effect of head teacher experience on student mental health.

Teaching is a relatively stable occupation in China, so teacher experience is closely related to age and, to a lesser extent, educational attainment. As shown in Supporting Information S1: Appendix Table A2, head teacher experience is strongly positively correlated with age and negatively correlated with years of schooling. Importantly, our identification strategy relies on simultaneously controlling for age and education. Without controls for age, the estimated effect of experience conflates learning‐by‐doing with life‐cycle effects. Age reflects offsetting influences, such as declining stamina, increasing professional maturity, and varying cultural proximity to students and parents, creating ambiguous net effects. Without controls for education, omitted variable bias arises because education and experience are negatively correlated in our sample (teachers with advanced degrees enter the profession later) yet both positively affect student mental health, attenuating the experience estimate toward zero. In China, most teachers' educational credentials are established before or shortly after labor market entry, whereas experience accumulates continuously throughout careers. Moreover, several institutional features generate considerable within‐group variation in teaching experience among teachers with similar education and age. Teachers enter the profession at different ages depending on whether they pursue advanced degrees or work in other sectors before teaching. Career interruptions, including maternity leave, caregiving responsibilities, sick leave, and temporary administrative reassignments, are common in the Chinese public sector and vary substantially in duration. Controlling for age and education, our baseline specification exploits this within‐group variation, isolating the effect of accumulated teaching practice while holding constant life‐cycle factors (age) and formal credentials (education). As a result, this specification yields a conservative estimate of the return to teaching experience, the policy‐relevant parameter for evaluating retention and professional development interventions. ¹⁵

4.2. Validity of Randomization

Our identification strategy relies on the assumption that students and teachers are plausibly randomly assigned to classes. We provide empirical evidence to support this assumption. First, following Gong et al. (2021) and Chung and Zou (2023), we implement a permutation test based on a resampling procedure to assess the randomness of student assignment. Specifically, we begin by computing the actual class‐level means for a set of predetermined student characteristics. For each class, we then pool students within the same school and randomly reassign them to hypothetical classes 10,000 times, maintaining fixed class sizes of 50 students and using resampling without replacement. ¹⁶ For each characteristic, we compare the actual class‐level mean to the distribution of simulated means and compute a p‐value as the proportion of simulations in which the simulated mean is less than the actual mean. If student assignment is truly random, the actual means should closely align with the simulated distribution. Therefore, a low p‐value (below 0.1) would indicate significant deviation and suggest non‐random sorting. As shown in Table 2, all p‐values exceed 0.1, thereby supporting the validity of the random assignment assumption for students in our sample.

TABLE 2.

Permutation test.

	(1)
Age	0.481
Male	0.492
Number of siblings	0.476
Agricultural hukou	0.478
Number of grades repeated in primary school	0.379
Number of grades skipped in primary school	0.180
Kindergarten attendance	0.499
Father's years of schooling	0.485
Mother's years of schooling	0.492

Note: Data are from the 2013–2014 CEPS. Column 1 reports p‐values from permutation tests of random assignment of students to classrooms within schools, based on predetermined student characteristics. For each characteristic, we compute the actual class‐level mean and compare it with the distribution of simulated means generated by 10,000 random reassignments of students within the same school, holding class size fixed at 50 and sampling without replacement. The reported p‐value is the proportion of simulations in which the simulated mean is less than the actual mean.

Second, we assess random assignment by testing whether teacher characteristics predict students' predetermined characteristics. We regress each predetermined student characteristic on head teacher characteristics, controlling for school fixed effects. Under random assignment, these variables should be uncorrelated. As shown in Supporting Information S1: Appendix Table A4, only 9 of the 90 estimated coefficients are statistically significant. This evidence suggests little systematic matching between head teacher characteristics and students' predetermined characteristics, supporting the random‐assignment assumption. Combined with the permutation tests in Table 2, which assess random assignment across students, these results provide consistent evidence that head teachers and students are randomly assigned in our sample.

5. Results

5.1. Baseline Results

In this section, we examine the effect of head teacher experience on student mental health. We begin by exploring the correlation between student mental health and head teacher experience. Figure 3 illustrates this pattern by plotting the residuals of the student mental health index against the residuals of head teacher experience. ¹⁷ The figure reveals that students assigned to more experienced head teachers tend to report better mental health outcomes.

FIGURE 3.

Correlation between the residual of student mental health and the residual of head teacher experience. Data are from the 2013–2014 CEPS. The figure presents the correlation between the residual of student mental health and the residual of the head teacher experience. The residuals are derived from regressions of the corresponding variables on school fixed effects. The dots represent the binned averages across 20 quantiles. The line represents linear fits.

We then estimate the causal effect of head teacher experience on student mental health using Equation (1). Table 3 presents the regression results, with all specifications controlling for head teacher and student characteristics as well as school fixed effects. ¹⁸ The results in column 1 show that each additional year of head teacher experience increases the average index of mental health by approximately 0.03 standard deviations. This corresponds to a 0.024‐point increase in the original average mental health score (i.e., the score before standardization). ¹⁹ Given that the mean of the unstandardized score in our sample is 4, the estimated effect represents a 0.6% improvement in student mental health for each additional year of head teacher experience.

TABLE 3.

The effects of head teacher experience on students' mental health.

	Average index of mental health	PCA index of mental health	Blue	Unhappy	Perceiving life as meaningless	Sad	Depressed
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Head teacher experience	0.027 a	0.026 a	0.017	0.021 a	0.038 b	0.019 c	0.014 c
	(0.011)	(0.011)	(0.010)	(0.010)	(0.010)	(0.010)	(0.008)
Student controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Head teacher controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes
School fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes
R 2	0.018	0.018	0.010	0.016	0.018	0.017	0.022
Observations	3529	3529	3529	3529	3529	3529	3529

Note: Data are from the 2013–2014 CEPS. Each column reports the results from a separate regression. Student controls include students' age, gender, number of siblings, number of grades repeated in primary school, number of grades skipped in primary school, parental years of schooling, and binary indicators for agricultural hukou and kindergarten attendance. Head teacher controls include head teachers' age, age squared, gender, years of schooling, professional title, and indicators for whether the head teacher teaches one of the core academic subjects, holds a tenured position, occupies an administrative position, has previous experience working at other schools, and teaches a class with more than 50 students. All regressions include school fixed effects. Standard errors in the parentheses are clustered at the school level.

^a significant at 5%.

^b significant at 1%.

^c significant at 10%.

In column 2, using the PCA index of mental health as an alternative outcome measure, we find that a 1‐year increase in head teacher experience improves the PCA index by approximately 0.03 standard deviations. ²⁰ Columns 3–7 examine the effect of head teacher experience on each of the five individual mental health items: feeling blue, unhappy, perceiving life as meaningless, sad, and depressed. ²¹ Most of the coefficients are statistically significant and similar in magnitude to those for the composite indices. Specifically, the results suggest that greater head teacher experience significantly reduces the likelihood that students report feeling unhappy, perceiving life as meaningless, sad, and depressed.

Supporting Information S1: Appendix Table A7 further reports our main specification and includes coefficients for head teacher characteristics and student controls. Head teacher age exhibits an inverted U‐shaped relationship with student mental health. The linear age term is positive, while the quadratic term is negative and statistically significant, suggesting diminishing marginal benefits of age at later career stages. The concave profile is consistent with life‐cycle factors. While younger teachers may benefit from increasing maturity and professional authority, these gains potentially level off or decline as teachers approach retirement. Such a pattern may be driven by age‐related declines in energy, a widening generational gap between teachers and students, or diminishing familiarity with contemporary adolescent challenges. Head teacher education (years of schooling) is also positively associated with student mental health. This pattern suggests that formal education and on‐the‐job experience capture distinct dimensions of teacher human capital and contribute independently to teacher effectiveness, implying complementarity rather than substitutability. This finding aligns with Qin et al. (2025), who find that one additional year of teacher schooling improves standardized CES‐D scores by 0.13 standard deviations. Our estimates in Supporting Information S1: Appendix Table A7 indicate that one additional year of schooling corresponds to improvements of approximately 0.08 standard deviations in the average and PCA index of mental health. Although our point estimate is slightly smaller than that reported by Qin et al. (2025), it is qualitatively consistent and falls within a plausible range, given differences in outcome measurement, sample composition, and model specification. Importantly, Qin et al. (2025) do not control for teacher experience in their baseline specifications. By jointly controlling for experience, age and education, our specification distinguishes learning‐by‐doing effects from life‐cycle factors and formal credentials. This decomposition provides complementary insights into distinct, policy‐relevant dimensions of teacher human capital.

5.2. Additional Analysis

This subsection presents two additional analyses. First, Table 3 estimates a linear relationship between teacher experience and student mental health. However, the literature presents mixed evidence regarding the returns to experience at different stages of a teaching career. Some studies suggest that the benefits of experience are concentrated in the early years (Rockoff 2004), while others report continued gains well into later stages (Chetty et al. 2011; Wiswall ²⁰¹³; Papay and Kraft ²⁰¹⁵). Following Huang et al. (2023), we categorize head teacher experience into six 5‐year intervals, with head teachers having five or fewer years of experience serving as the reference group. The results of Table 4 show that the positive association between head teacher experience and student mental health persists beyond the early career stage, with no evidence of sharp declines at higher experience levels. This pattern is consistent with learning‐by‐doing in the head teacher role. Importantly, our specification controls for age and age squared, allowing us to separate experience effects (accumulated pastoral care skills) from life‐cycle effects (maturity, energy, generational proximity). Any age‐related influences, such as the inverted U‐shaped relationship between teacher age and effectiveness documented in Supporting Information S1: Appendix Table A7, are absorbed by the age controls rather than confounding the experience estimates. ²²

TABLE 4.

The nonlinear effects of head teacher experience on students' mental health.

	Average index of mental health	PCA index of mental health
	(1)	(2)
Head teacher experience (6–10 years)	0.225	0.225
	(0.145)	(0.145)
Head teacher experience (11–15 years)	0.460 a	0.459 a
	(0.180)	(0.179)
Head teacher experience (16–20 years)	0.632 a	0.631 a
	(0.311)	(0.310)
Head teacher experience (21–25 years)	0.715 a	0.713 a
	(0.300)	(0.298)
Head teacher experience (above 26 years)	0.639	0.633
	(0.400)	(0.398)
Student controls	Yes	Yes
Head teacher controls	Yes	Yes
School fixed effects	Yes	Yes
R 2	0.019	0.019
Observations	3529	3529

Note: Data are from the 2013–2014 CEPS. We categorize head teacher experience into six 5‐year intervals, with head teachers having five or fewer years of experience serving as the reference group. Each column reports the results from a separate regression. Student controls include students' age, gender, number of siblings, number of grades repeated in primary school, number of grades skipped in primary school, parental years of schooling, and binary indicators for agricultural hukou and kindergarten attendance. Head teacher controls include head teachers' age, age squared, gender, years of schooling, professional title, and indicators for whether the head teacher teaches one of the core academic subjects, holds a tenured position, occupies an administrative position, has previous experience working at other schools, and teaches a class with more than 50 students. All regressions include school fixed effects. Standard errors in the parentheses are clustered at the school level.

^a significant at 5%.

To further assess potential nonlinearity, we re‐estimate the baseline specification including linear and quadratic terms in head teacher experience. This flexible specification allows the marginal effect of experience to vary continuously across the experience distribution and provides a direct test for diminishing or negative returns at higher experience levels. As shown in Supporting Information S1: Appendix Table A8, the quadratic term is small and statistically insignificant, while the linear term remains positive. We find no evidence of a concave or inverted U‐shaped relationship between teaching experience and effectiveness. Instead, the results reinforce the conclusion that the benefits of experience persist throughout teachers' careers, consistent with Gerritsen et al. (2017) and Huang et al. (2023).

Second, to assess the persistence of these effects over time, we use follow‐up data from the 2014–2015 wave of CEPS, which tracks the same cohort of 7th‐grade students one year later. This wave includes 10 items that closely correspond to the CES‐D scale, asking about the frequency of specific feelings during the past seven days on a scale from 1 (never) to 5 (always). These items include: (1) blue, (2) depressed, (3) unhappy, (4) perceiving life as meaningless, (5) having no passion to do anything, (6) sad, (7) nervous, (8) worried, (9) feeling something bad will happen, and (10) too energetic to concentrate in class. ²³ We assess internal consistency using Cronbach's alpha. The 10‐item mental health scale exhibits a Cronbach's alpha of 0.92, indicating high reliability. We recode these items so that higher values correspond to better mental health and construct average and PCA indices using the same methodology as our baseline analysis. As shown in Table 5, the estimated medium‐term effects of Grade 7 head teacher experience on Grade 8 student mental health remain positive and statistically significant. This demonstrates that our findings are robust to using a more comprehensive mental health measure, assessing outcomes 1 year after initial exposure, and using different item wordings.

TABLE 5.

The medium‐term effects of head teacher experience on Grade 8 students' mental health.

	Average index of mental health	PCA index of mental health
	(1)	(2)
Head teacher experience	0.016 a	0.021 b
	(0.009)	(0.009)
Student controls	Yes	Yes
Head teacher controls	Yes	Yes
School fixed effects	Yes	Yes
R 2	0.018	0.016
Observations	3230	3230

Note: Data are from the 2013–2014 and the 2014–2015 CEPS. Each column reports the results from a separate regression. Student controls include students' age, gender, number of siblings, number of grades repeated in primary school, number of grades skipped in primary school, parental years of schooling, and binary indicators for agricultural hukou and kindergarten attendance. Head teacher controls include head teachers' age, age squared, gender, years of schooling, professional title, and indicators for whether the head teacher teaches one of the core academic subjects, holds a tenured position, occupies an administrative position, has previous experience working at other schools, and teaches a class with more than 50 students. All regressions include school fixed effects. Standard errors in the parentheses are clustered at the school level.

^a significant at 10%.

^b significant at 5%.

5.3. Heterogeneity Analysis

The baseline results demonstrate a positive effect of head teacher experience on students' mental health. We further explore potential heterogeneity in this effect across different subgroups, including gender, family socioeconomic status (SES), and school location.

First, prior research finds that girls are more prone to internalizing emotional problems than boys, beginning in early adolescence and continuing into adulthood (Hankin and Abramson 1999; Giulietti et al. ²⁰²²). Motivated by this evidence, we investigate whether the effect of head teacher experience on students' mental health varies by gender. Specifically, we interact head teacher experience with a male indicator in Equation (1). The results in Panel A of Table 6 show no statistically significant difference in the effect between male and female students. This finding suggests that the mental health benefits associated with more experienced head teachers are broadly consistent across genders.

TABLE 6.

The heterogeneous effects of head teacher experience on students' mental health.

	Average index of mental health	PCA index of mental health
	(1)	(2)
Panel A. Gender
Head teacher experience × Male	−0.003	−0.003
	(0.006)	(0.006)
Head teacher experience	0.028 a	0.028 a
	(0.011)	(0.011)
R 2	0.019	0.019
Panel B. Family SES
Head teacher experience × Low‐SES	0.009 b	0.009 b
	(0.005)	(0.005)
Head teacher experience	0.020 b	0.019
	(0.012)	(0.012)
R 2	0.019	0.019
Panel C. School location
Head teacher experience × Rural	0.022 a	0.021 a
	(0.010)	(0.010)
Head teacher experience	0.013	0.012
	(0.014)	(0.014)
R 2	0.020	0.020
Student controls	Yes	Yes
Head teacher controls	Yes	Yes
School fixed effects	Yes	Yes
Observations	3529	3529

Note: Data are from the 2013–2014 CEPS. Each column reports the results from a separate regression. Student controls include students' age, gender, number of siblings, number of grades repeated in primary school, number of grades skipped in primary school, parental years of schooling, and binary indicators for agricultural hukou and kindergarten attendance. Head teacher controls include head teachers' age, age squared, gender, years of schooling, professional title, and indicators for whether the head teacher teaches one of the core academic subjects, holds a tenured position, occupies an administrative position, has previous experience working at other schools, and teaches a class with more than 50 students. All regressions include school fixed effects. Standard errors in the parentheses are clustered at the school level.

^a significant at 5%.

^b significant at 10%.

Second, we explore heterogeneity by family socioeconomic status, a well‐established determinant of adolescent well‐being (Akee et al. 2018). We use parental education as a proxy for SES, classifying students as low‐SES if both parents have attained no more than a senior high school education. The results in Panel B of Table 6 indicate that the positive effect of head teacher experience is significantly larger for students from low‐SES backgrounds. This finding underscores the particularly important role that experienced teachers play in supporting the mental health of socioeconomically disadvantaged students and suggests that teacher quality may help mitigate mental health disparities. One plausible explanation is that experienced teachers are better equipped to compensate for the limited educational and emotional support often found in low‐SES households.

Finally, we examine whether the effect of head teacher experience differs by school location. The results in Panel C of Table 6 indicate that the positive impact on mental health is more pronounced for students attending rural schools. Since rural students are more likely to come from economically disadvantaged families, this finding provides further evidence that the benefits of head teacher experience are especially salient for students from lower‐SES backgrounds.

5.4. The Impact of Subject Teachers

In Section 2, we introduce the institutional context of China's educational system, where head teachers play a central role in students' daily lives. This raises the question of whether the observed effects of teacher experience on mental health are specific to head teachers or reflect a broader influence of teaching experience more generally. Therefore, we examine whether the experience of subject teachers, who are not head teachers, affects student mental health.

We construct a measure of subject teacher experience as the average years of teaching among core subject teachers who are not head teachers in a given class. ^{24 , 25} The results in Table 7 show that subject teacher experience has no statistically significant effect on student mental health. This divergence in effects aligns with the differentiated roles occupied by head and subject teachers. Subject teachers are primarily responsible for instructing a specific course and ensuring students meet subject‐specific academic goals. In contrast, head teachers assume a broader managerial and pastoral role, often focusing on the development of students' general skills and mental health. These findings suggest that the mental health benefits of teacher experience are driven specifically by the unique responsibilities and sustained engagement of head teachers.

TABLE 7.

The effects of subject teacher experience on students' mental health.

	Average index of mental health	PCA index of mental health
	(1)	(2)
Subject teacher experience	0.008	0.009
	(0.017)	(0.017)
Student controls	Yes	Yes
Subject teacher controls	Yes	Yes
School fixed effects	Yes	Yes
R 2	0.019	0.019
Observations	3529	3529

Note: Data are from the 2013–2014 CEPS. Each column reports the results from a separate regression. Student controls include students' age, gender, number of siblings, number of grades repeated in primary school, number of grades skipped in primary school, parental years of schooling, and binary indicators for agricultural hukou and kindergarten attendance. Subject teacher controls include subject teachers' age, age squared, gender, years of schooling, professional title, and indicators for whether the subject teacher holds a tenured position, occupies an administrative position, has previous experience working at other schools, and teaches a class with more than 50 students. All regressions include school fixed effects. Standard errors in the parentheses are clustered at the school level.

To further compare the effects of subject teacher experience with those of head teacher experience, we include both variables in the same regression specification. The results in Supporting Information S1: Appendix Table A9 show that head teacher experience continues to exert a statistically significant positive effect on student mental health, whereas subject teacher experience remains insignificant.

6. Mechanism Analysis

This section investigates the potential mechanisms through which head teacher experience affects student mental health, focusing on several channels: teacher–student relationships, teacher–parent interactions, teacher work satisfaction, peer interactions, teachers' behavioral responses to students' mental health problems, and academic performance. ²⁶

6.1. Teacher–Student Relationships

Teacher–student relationships are a critical determinant of student mental health. Prior research highlights that mental health is closely linked to the care, attention, and support students receive from their teachers, with strong relationships fostering a sense of safety and belonging, which in turn enhances mental health outcomes (Lavy and Schlosser 2011; Novak and Kawachi ²⁰¹⁵). We hypothesize that more experienced head teachers are better equipped to cultivate supportive relationships with students, thereby reducing mental health problems.

To assess the quality of teacher–student relationships, we incorporate three measures: two based on parental perceptions and one from teacher reports. First, parents are asked, “Does this child like his/her head teachers at school?” with responses recorded on a four‐point scale from 1 (“strongly dislike”) to 4 (“like very much”). We construct a binary variable, student affection toward the head teacher, coded as 1 if the response is 4, and 0 otherwise. Second, parents are asked: “Do you think the teacher is responsible and patient with this child?” Responses are recorded on a five‐point scale from 1 (“not at all”) to 5 (“very much”). We define a binary variable, teacher responsibility and patience, equal to 1 if parents answer that the teacher is “very much” responsible and patient with the child, and 0 otherwise. Finally, we use the variable perceived student respect, derived from the head teacher questionnaire. Head teachers are asked: “What proportion of students in your class do you feel respect you?” Responses are recorded on a five‐point scale ranging from 1 (“very few”) to 5 (“almost all”). We construct a binary variable equal to 1 if the response is 5, and 0 otherwise.

The results in Table 8 indicate that head teacher experience significantly strengthens teacher–student relationships. Specifically, greater experience increases the likelihood that students express affection toward their head teacher, that parents provide more favorable evaluations of the teacher's responsibility and patience, and that students report higher levels of respect for the head teacher. These findings suggest that experienced head teachers cultivate more supportive teacher–student relationships, which in turn enhances student mental health.

TABLE 8.

The effects of head teacher experience on teacher–student relationships.

	Student affection toward the head teacher	Teacher responsibility and patience	Perceived student respect
	(1)	(2)	(3)
Head teacher experience	0.012 a	0.010 b	0.066 c
	(0.005)	(0.006)	(0.024)
Student controls	Yes	Yes	Yes
Head teacher controls	Yes	Yes	Yes
School fixed effects	Yes	Yes	Yes
R 2	0.071	0.080	0.716
Observations	3392	3400	3454

Note: Data are from the 2013–2014 CEPS. Each column reports the results from a separate regression. Student controls include students' age, gender, number of siblings, number of grades repeated in primary school, number of grades skipped in primary school, parental years of schooling, and binary indicators for agricultural hukou and kindergarten attendance. Head teacher controls include head teachers' age, age squared, gender, years of schooling, professional title, and indicators for whether the head teacher teaches one of the core academic subjects, holds a tenured position, occupies an administrative position, has previous experience working at other schools, and teaches a class with more than 50 students. All regressions include school fixed effects. Standard errors in the parentheses are clustered at the school level.

^a significant at 5%.

^b significant at 10%.

^c significant at 1%.

6.2. Teacher–Parent Interactions

Head teachers frequently engage with parents regarding students' academic performance and general well‐being, often offering guidance on supporting child development. ²⁷ Prior research suggests that these teacher–parent interactions, particularly those involving the exchange of information, can improve student performance (Avvisati et al. 2014). More experienced head teachers, who are likely to possess stronger communication skills and a deeper understanding of student development, may be better equipped to initiate and foster constructive dialog with parents. Increased engagement with head teachers may enhance parental understanding of their children's emotional and psychological needs, thereby contributing to improved mental health. Motivated by this insight, we examine whether teacher–parent interaction serves as a potential mechanism through which head teacher experience influences student mental health.

To measure teacher–parent interaction, we use responses from parent questionnaires, focusing on the substance and frequency of communication. From the parent questionnaire, parents are asked: “When contacting the child's teacher this semester, what topics were usually discussed?” If the response includes “schoolwork” “mental health” or “physical health”, we define a binary variable, parent–teacher discussion of school life, equal to 1. Similarly, to capture teacher‐initiated contact, parents are asked: “When the teacher contacted parents this semester, what did he/she usually talk about with parents?” If the reported topic is “schoolwork” “mental health” or “physical health” during these interactions, we define a binary variable, teacher–parent discussion of school life, coded as 1. Furthermore, we construct a variable for parental supervision. This is based on the question: “To what extent could you meet the requirement this semester if the teachers require the parents to check their child's homework, such as writing checked, signing names, etc.?” Responses range from 1 (“the teachers do not have such requirement”) to 5 (“not at all”). We code parental supervision as 1 if parents indicate compliance (responses of 2 or lower, representing “completely” meeting requirements or no requirement). Finally, we define parent meeting attendance using responses to the question: “Do you attend the parent meeting this semester”. This variable is set to 1 if parents respond “the school held one, and the parents did attend” or “not held yet, but the parents are going to attend once held.” It is set to 0 if parents respond “the school held one, but the parents didn't attend” or “not held yet, and the parents are not going to attend if held.”

The results in Table 9 indicate that the likelihood of parent–teacher discussions concerning school life and parental supervision significantly increases with head teacher experience. These strengthened interactions likely contribute to improved student mental health by enhancing parental awareness and promoting more effective collaboration between parents and teachers in addressing students' academic and emotional needs.

TABLE 9.

The effects of head teacher experience on teacher–parent interactions.

	Parent–teacher discussion of school life	Teacher–parent discussion of school life	Parental supervision	Parent meeting attendance
	(1)	(2)	(3)	(4)
Head teacher experience	0.004 a	0.006 a	0.013 b	0.004
	(0.003)	(0.003)	(0.006)	(0.003)
Student controls	Yes	Yes	Yes	Yes
Head teacher controls	Yes	Yes	Yes	Yes
School fixed effects	Yes	Yes	Yes	Yes
R 2	0.061	0.089	0.061	0.148
Observations	3379	3351	3387	3415

Note: Data are from the 2013–2014 CEPS. Each column reports the results from a separate regression. Student controls include students' age, gender, number of siblings, number of grades repeated in primary school, number of grades skipped in primary school, parental years of schooling, and binary indicators for agricultural hukou and kindergarten attendance. Head teacher controls include head teachers' age, age squared, gender, years of schooling, professional title, and indicators for whether the head teacher teaches one of the core academic subjects, holds a tenured position, occupies an administrative position, has previous experience working at other schools, and teaches a class with more than 50 students. All regressions include school fixed effects. Standard errors in the parentheses are clustered at the school level.

^a significant at 10%.

^b significant at 5%.

6.3. Teachers' Work Satisfaction

Existing research highlights the importance of teacher job satisfaction and emotional well‐being in shaping student mental health outcomes (Jennings and Greenberg 2009). One potential mechanism through which head teacher experience influences student mental health is by improving teacher job satisfaction and reducing emotional exhaustion. Greater experience may strengthen professional identity, clarify role expectations, and enhance teachers' sense of efficacy. These factors promote greater psychological stability among teachers and, in turn, support better student mental health.

To measure work‐related fatigue, the CEPS asks head teachers: “Have you ever felt tired of being a teacher?” Responses are recorded on a four‐point scale from 1 (“never”) to 4 (“often”), from which we define the variable emotional exhaustion. The CEPS also measures overall job satisfaction by asking head teachers: “In general, are you satisfied with your current job as a head teacher?” Responses are recorded on a five‐point scale ranging from 1 (“strongly unsatisfied”) to 5 (“strongly satisfied”). In addition, head teachers are asked whether, given a second chance, they would “choose to become a junior‐high teacher again” and “choose to continue teaching at their current school”. Responses to both questions are recorded on a three‐point scale from 1 (“no”) to 3 (“definitely yes”), which we use to construct the variables willingness to be a teacher again and willingness to continue teaching at the current school.

The results in Table 10 show that head teachers with more years of experience exhibit significantly lower levels of emotional exhaustion, as well as a stronger willingness to remain in the teaching profession and to continue working at their current school. These findings suggest that teaching experience enhances occupational well‐being, which may, in turn, positively affect student mental health. Teachers who are more satisfied with their jobs and experience less emotional fatigue are more likely to engage meaningfully with students, provide emotional support, and remain responsive to their mental health needs.

TABLE 10.

The effects of head teacher experience on teacher's work satisfaction.

	Emotional exhaustion	Job satisfaction	Willingness to be a teacher again	Willingness to continue teaching at the current school
	(1)	(2)	(3)	(4)
Head teacher experience	−0.076 a	0.035	0.040 b	0.040 b
	(0.037)	(0.039)	(0.023)	(0.020)
Student controls	Yes	Yes	Yes	Yes
Head teacher controls	Yes	Yes	Yes	Yes
School fixed effects	Yes	Yes	Yes	Yes
R 2	0.663	0.701	0.520	0.764
Observations	3505	3505	3505	3505

Note: Data are from the 2013–2014 CEPS. Each column reports the results from a separate regression. Student controls include students' age, gender, number of siblings, number of grades repeated in primary school, number of grades skipped in primary school, parental years of schooling, and binary indicators for agricultural hukou and kindergarten attendance. Head teacher controls include head teachers' age, age squared, gender, years of schooling, professional title, and indicators for whether the head teacher teaches one of the core academic subjects, holds a tenured position, occupies an administrative position, has previous experience working at other schools, and teaches a class with more than 50 students. All regressions include school fixed effects. Standard errors in the parentheses are clustered at the school level.

^a significant at 5%.

^b significant at 10%.

6.4. Peer Effects

A growing body of literature establishes a direct link between school peers and students' mental health (Giulietti et al. 2022; Kiessling and Norris ²⁰²²; Chen et al. ²⁰²⁶). Because adolescence is a period of heightened sensitivity to peer relationships, and the classroom is a primary setting for daily social interaction, head teachers may shape student mental health indirectly by managing classroom climate and peer dynamics. Head teachers influence peer interactions by setting and enforcing norms of respect and inclusion, monitoring behavior to deter bullying, and fostering peer bonding through group activities. These actions may improve student mental health. If experienced head teachers perform these tasks more effectively, they may cultivate more supportive peer environments, with greater inclusion and less conflict, providing a plausible channel through which head teacher experience improves student mental health.

We examine peer effects as a potential mechanism linking head teacher experience to student mental health. The 2013–2014 CEPS baseline wave asks students to report their agreement with the following statements about school life: (1) my class is in a good atmosphere and (2) most of my classmates are nice to me, rated on a four‐point scale from 1 (strongly disagree) to 4 (strongly agree). We construct indicator variables equal to one if the student reports “strongly agree” or “somewhat agree,” and zero otherwise. The CEPS also asks students how frequently they engage in activities with classmates, including (1) visiting museums, zoos, or science museums and (2) going out to watch movies, shows, or sports games. Responses are rated on a five‐point scale ranging from 1 (never) to 5 (more than once a month). For these items, we define indicator variables equal to one if the reported frequency is “once a month” or “more than once a month.” These measures capture perceived peer climate (atmosphere, kindness) and behavioral manifestations of peer relationships (voluntary social activities outside school).

As shown in Table 11, head teacher experience is positively associated with several dimensions of the peer environment. Students taught by more experienced head teachers are more likely to report a positive classroom atmosphere. They also report participating more frequently in cultural activities and entertainment events with classmates. These results are consistent with learning‐by‐doing. As teachers accumulate experience, they develop the specialized skills to create positive classroom atmospheres where students perceive peers as kind and supportive. Moreover, they develop strategies for facilitating peer bonding, as evidenced by the higher frequency with which their students engage in cultural and entertainment activities together outside school. These patterns suggest that improved peer interactions are a plausible channel through which head teacher experience enhances student mental health.

TABLE 11.

The effects of head teacher experience on classroom environment and peer interactions.

	Good classroom atmosphere	Classmates are friendly	Visiting museums, zoos, science museums	Going out to watch movies, shows, sports games
	(1)	(2)	(3)	(4)
Head teacher experience	0.010 a	0.001	0.005 b	0.006 a
	(0.006)	(0.002)	(0.002)	(0.003)
Student controls	Yes	Yes	Yes	Yes
Head teacher controls	Yes	Yes	Yes	Yes
School fixed effects	Yes	Yes	Yes	Yes
R 2	0.094	0.047	0.078	0.106
Observations	3504	3497	3418	3408

Note: Data are from the 2013–2014 CEPS. Each column reports the results from a separate regression. Student controls include students' age, gender, number of siblings, number of grades repeated in primary school, number of grades skipped in primary school, parental years of schooling, and binary indicators for agricultural hukou and kindergarten attendance. Head teacher controls include head teachers' age, age squared, gender, years of schooling, professional title, and indicators for whether the head teacher teaches one of the core academic subjects, holds a tenured position, occupies an administrative position, has previous experience working at other schools, and teaches a class with more than 50 students. All regressions include school fixed effects. Standard errors in the parentheses are clustered at the school level.

^a significant at 10%.

^b significant at 5%.

6.5. Teachers' Behavioral Responses to Students' Mental Health Problems

Teachers' behavioral responses to students' mental health problems may, in turn, affect students' mental health. Therefore, we examine whether head teacher experience affects teachers' responses when students experience mental health problems. The 2013–2014 CEPS baseline wave does not include detailed measures of teachers' responses when students experience mental health problems, which prevents us from directly testing this mechanism in our main sample. However, the 2014–2015 CEPS follow‐up wave contains rich information on teacher behaviors in such situations, allowing us to examine this channel using data from the same student cohort one year later. Specifically, students indicate on a four‐point scale ranging from 1 (“strongly disagree”) to 4 (“strongly agree”) whether the following statements apply when they experience mental health problems or negative emotions: (1) the teacher ignores or isolates me; (2) the teacher criticizes me in public; (3) the teacher criticizes me in private; (4) the teacher tries to help me; and (5) the teacher asks my parents to help address the problem. To construct the composite index, we reverse‐code the three negative items (ignoring/isolating, public criticism, private criticism) so that higher values indicate more supportive responses. We then aggregate these five items and normalize the resulting sum within each school to have a mean of 0 and a standard deviation of 1, yielding a standardized measure of supportive teacher behavior.

We examine whether head teacher experience influences these behavioral responses by re‐estimating our baseline specifications using the 2014–2015 teacher‐behavior measures as dependent variables. As shown in Supporting Information S1: Appendix Table A10, head teacher experience is associated with a significantly lower likelihood of ignoring or isolating students and a significantly higher likelihood of attempting to help them when mental health problems arise. Moreover, head teacher experience is positively associated with the composite index of supportive teacher behavior. This pattern reflects learning‐by‐doing. Through repeated exposure to diverse student situations, experienced head teachers develop superior emotional intelligence and mental health literacy, improved emotional regulation, and a broader response repertoire. These results suggest that teacher behaviors in response to students' mental health problems are a plausible mechanism linking head teacher experience to better student mental health outcomes. However, the behavioral measures were collected in the 2014–2015 wave (Grade 8), whereas our primary mental health outcomes are from the 2013–2014 baseline (Grade 7). Because the mediator is measured after the outcome, we cannot establish formal mediation or quantify the contribution of this channel through standard decomposition methods. Accordingly, we interpret the findings as suggestive evidence on mechanisms rather than definitive proof that these behaviors mediate the relationship between head teacher experience and student mental health.

6.6. Decomposition Analysis

To quantify the relative contribution of each mechanism underlying the effect of head teacher experience on student mental health, we employ the decomposition method proposed by Gelbach (2016) and Gong et al. (2018). Specifically, let $m_{ics}^j$ represent mechanism variable j, and estimate each channel using the following specification:

$$m_{ics}^j={\zeta }_0^j+{\zeta }_1^j{TeacherExp}_{cs}+{X_{ics}}^{\mathit{\prime }}\mathrm{\Gamma }+{Z_{cs}}^{\mathit{\prime }}\mathrm{\Pi }+{\phi }_s+{\epsilon }_{ics}^j$$ (2)

We then extend the baseline specification by including all mechanism variables:

$$Y_{ics}={\alpha }^{\prime }+{\beta }^{\prime }{TeacherExp}_{cs}+\sum _j{\phi }^j{m}_{ics}^j+{X_{ics}}^{\mathit{\prime }}\mathrm{\Gamma }+{Z_{cs}}^{\mathit{\prime }}\mathrm{\Pi }+{\phi }_s+{\epsilon }_{ics}$$ (3)

As shown by Gelbach (2016), the total effect of head teacher experience $\hat{\beta }$, can be decomposed into the portion explained by the included mechanisms and the unexplained part:

$$\hat{\beta }=\widehat{{\beta }^{\prime }}+\sum _j{\hat{\phi }}^j{\hat{\zeta }}_1^j$$ (4)

This framework implies that the explanatory power of each mechanism j is given by the proportion ${\hat{\phi }}^j{\hat{\zeta }}_1^j/\hat{\beta }$, indicating its relative contribution to the total effect of head teacher experience on student mental health.

Figure 4 presents the results of the decomposition analysis for the effect of head teacher experience on the average mental health index and the PCA mental health index. ²⁸ For the average index, we find that teacher–student relationships account for 29.9% of the total effect, teacher–parent interactions for 11.8%, teacher's work satisfaction for 12.2%, and peer effects for 12.1%. These four mechanisms explain approximately 66% of the total effect of head teacher experience on the average index of mental health. The decomposition analysis for the PCA mental health index yields similar results. Teacher–student relationships account for 31.7%, teacher–parent interactions for 12.3%, Teacher's work satisfaction for 12.8%, and peer effects for 12.5% of the total effect.

FIGURE 4.

Decomposition of head teacher experience effects by channels: Average index of mental health and PCA index of mental health. Data are from the 2013–2014 CEPS. The figure presents the decomposition of the effects of head teacher experience on students' mental health outcomes. The effects on average index of mental health can be explained by teacher–student relationships (29.9%), teacher–parent interactions (11.8%), teacher's work satisfaction (12.2%), peer effects (12.1%), and other factors. The effects on the PCA index of mental health can be explained by teacher–student relationships (31.7%), teacher–parent interactions (12.3%), teacher's work satisfaction (12.8%), peer effects (12.5%), and other factors.

6.7. Alternative Mechanism: Academic Performance

Head teacher experience may influence student mental health through academic performance, given evidence that educational achievement can improve students' mental health (Weng 2025). We investigate this channel using two measures of academic outcomes. First, we use test scores from three core subjects (Chinese, Mathematics, and English), which are provided by the students' respective school administration offices from the midterm examination in Grade 7. We aggregate the test scores across the three core subjects and standardize the resulting sum to have a mean of 0 and a standard deviation of 1 within each school, ensuring comparability across schools with different grading standards. Second, CEPS asks students to assess the difficulty they experience in learning each core subject on a four‐point scale ranging from 1 (“very difficult”) to 4 (“easy”). We construct a binary indicator, perceived difficulty of core subjects, which equals 1 if the student reports at least one subject as “difficult” or “very difficult,” and 0 otherwise. This subjective measure captures academic confidence and self‐efficacy.

We assess whether head teacher experience affects students' academic performance by re‐estimating the baseline specifications with exam scores and perceived difficulty of core subjects as outcomes. As shown in Supporting Information S1: Appendix Table A11, head teacher experience has no statistically significant effect on objective exam scores or students' perceived difficulty of core subjects. To further examine whether academic performance mediates the relationship between head teacher experience and student mental health, we add exam scores and perceived difficulty of core subjects as control variables to our baseline regression with the mental health index as the dependent variable. The results in Supporting Information S1: Appendix Table A12 show that the estimated effect of head teacher experience on student mental health remains stable after conditioning on these academic achievement measures. This suggests that the relationship is not mediated by academic performance. This conclusion is consistent with evidence that returns to teacher experience are limited for achievement outcomes. e.g., in a review of U.S. evidence, Hanushek (2003) reports that 71% of surveyed estimates find no significant improvement, or even a decline, in academic performance associated with teacher experience. Similarly, Rivkin et al. (2005) show that observable teacher characteristics, including experience, explain a negligible share of the variation in student achievement gains.

7. Robustness Check

7.1. Tests for Sample Randomization

Random assignment of students and teachers to classes within schools is essential for identifying the causal effect of head teacher experience on students' mental health. Although we have previously validated the plausibility of this assumption, concerns may remain that some schools reporting random assignment did not implement it in practice. To address this potential threat to identification, we further assess the validity of our identifying assumption.

First, concerns regarding potential sorting may arise if schools systematically assign more experienced head teachers to classes with students exhibiting greater mental health challenges. To mitigate this concern, we follow the approach of Deryugina and Molitor (2020). We first regress student mental health on a set of predetermined student characteristics to construct predicted mental health values. We then regress head teacher experience on these predicted values. As shown in Supporting Information S1: Appendix Table A13, the estimated relationship between head teacher experience and predicted student mental health is statistically insignificant. This result suggests that sorting based on students' mental health is minimal, lending support to the assumption of random class assignment in our sample.

Second, following Gong et al. (2018), we conduct an empirical validation exercise. We randomly exclude a subset of schools from the estimation sample and assess whether the resulting estimates differ meaningfully from the baseline results. If most schools in the sample indeed assign students randomly to classrooms, the re‐estimated coefficients should remain close to the baseline estimates reported in Table 3. Our baseline sample includes 43 schools; in each iteration, we randomly drop 10% (i.e., 4 schools) and re‐estimate Equation (1), repeating this procedure 1000 times. This yields a distribution of 1000 estimated coefficients. Supporting Information S1: Appendix Figure A2 plots the distribution of these coefficients. Panel A shows that when the average index of mental health is used as the outcome variable, the coefficients are approximately normally distributed and centered around the baseline coefficient of 0.027. Panel B presents similar results for the PCA index of mental health. These findings suggest that the majority of schools in our sample follow random class assignment practices, supporting the validity of our identification strategy.

Third, we restrict the analysis to the sample of schools that do not employ random class assignment to estimate the effect of head teacher experience on student mental health. This exercise provides a benchmark for comparison with the randomized sample and allows us to assess the potential bias introduced by including non‐randomly assigned classes. The results in Supporting Information S1: Appendix Table A14 show that in these schools, head teacher experience has no statistically significant effect on student mental health, and the estimated coefficients are close to zero. This suggests that including non‐randomly assigned schools in the estimation sample would likely attenuate the estimated effects, leading to downward‐biased estimates.

7.2. Test for Omitted Variable Bias

While our estimation model controls for a set of observed teacher characteristics, our estimates may still reflect unobserved teacher characteristics rather than solely the effect of teacher experience. To address this, we employ the Oster (2019) method to evaluate the robustness of our main results to potential omitted variable bias from unobserved teacher characteristics. The key insight of this approach is that if adding control variables substantially increases the model's explanatory power (as measured by R ²) without materially changing the coefficient of interest, then the estimates are likely to be robust to omitted variable bias. We compute δ, which captures the relative importance of omitted variables compared to included controls. Following Oster's recommended benchmarks, we set R ² _max = 1.3 ${\tilde{R}}^2$, where ${\tilde{R}}^2$ denotes the R ² from the fully controlled specification of Equation (1). We then examine whether δ exceeds one under the assumption that β = 0 and R ² _max = 1.3 ${\tilde{R}}^2$. ²⁹ The results in Supporting Information S1: Appendix Table A15 show that δ is estimated to be 5.2 for the average index of mental health and 5.5 for the PCA index of mental health. These values imply that selection on unobservables would need to be more than five times stronger than selection on observables to fully attenuate the estimated effects. This magnitude of bias is implausibly large, suggesting that our results are unlikely to be driven by omitted variable bias.

7.3. Falsification Test

To assess whether our estimated effects are driven by chance, we conduct a falsification test by randomly reassigning head teacher experience across classes and re‐estimating its effect on students' mental health. Specifically, we randomly assign to each student within a class a head teacher experience value drawn from the full sample, while keeping all other control variables at their original levels. This procedure is repeated 1000 times, and Equation (1) is re‐estimated in each iteration using the simulated head teacher experience as the key explanatory variable. Supporting Information S1: Appendix Figure A3 presents the distribution of the estimated coefficients from these 1000 falsification regressions, using both the average index and PCA index of mental health. The coefficients are approximately normally distributed around zero, indicating that our main findings are unlikely to be driven by random variation.

7.4. Using Mental Health Thresholds to Measure Student Mental Health

In the baseline analysis, we measure student mental health using a standardized continuous index. In this subsection, we complement that approach by applying clinically motivated thresholds to classify students as experiencing clinically meaningful psychological distress. The 2013–2014 CEPS wave includes five mental health items, which we aggregate to construct a raw composite score ranging from 5 to 25. While there is no single universally accepted threshold for the MHI‐5, we follow the approach of Stillman et al. (2009), who examined the relationship between migration and mental health using the MHI‐5. We adopt two conservative thresholds, scores below 19 or below 20, to classify students as experiencing clinically meaningful psychological distress. These thresholds allow us to examine how head teacher experience affects the probability that students fall below clinically relevant levels. The results in Supporting Information S1: Appendix Table A16 show that head teacher experience has a statistically significant positive effect on the raw composite mental health score, consistent with our baseline results. Moreover, head teacher experience significantly reduces the probability of scoring below the 19 and 20 thresholds by 0.9 and 1.1% points, respectively. These results suggest a reduction in the risk of severe psychological distress.

7.5. Sample Attrition

Head teacher experience may influence students' likelihood of completing the mental health questionnaire. If students assigned to more experienced head teachers are systematically more willing to disclose information about their mental health, this could introduce bias into our estimates. To evaluate this concern, we examine whether head teacher experience predicts the probability of completing the mental health scale. We construct a binary variable, attrition dummy, which equals 1 if a student did not complete the mental health scale and 0 otherwise. As shown in Supporting Information S1: Appendix Table A17, head teacher experience is not significantly associated with the likelihood of completing the scale, suggesting that selective non‐response is unlikely to bias our results.

8. Conclusion

Addressing adolescent mental health is essential for supporting students' academic achievement and long‐term success in the labor market. Despite growing policy attention, relatively little is known about how the school environment affects student mental health. This study estimates the causal effect of teacher quality, measured by teacher experience, on the mental health of junior high school students in China. We find that each additional year of head teacher experience improves the average index and PCA index of mental health by 0.03 standard deviations. By contrast, no such effects are observed for subject teachers who are solely responsible for academic instruction.

The mechanism analysis indicates that the positive effect of head teacher experience on student mental health operates through several channels, including stronger teacher–student relationships, more frequent teacher–parent interactions, higher work satisfaction, and a more supportive peer interaction. We also find suggestive evidence that experienced head teachers provide more supportive responses to students' mental health problems. The pattern documented in our mechanism analysis is closely consistent with an authoritative pastoral approach: more experienced teachers combine high warmth, reflected in strong relationships, empathetic responses to student distress, and proactive communication with parents, with high structure, reflected in effective classroom management, inclusive norms, and the facilitation of positive peer interactions.

Our findings suggest that teacher experience yields substantial returns to adolescent mental health, providing an empirical basis for school‐based interventions. These results support policies designed to mitigate the avoidable attrition of senior head teachers. Such policies may include competitive compensation, workload relief, enhanced administrative support, and institutional recognition of the demanding pastoral care responsibilities inherent in the head teacher role. In addition, building on our mechanism analysis, we propose two primary policy directions. First, professional development for early‐career head teachers should be competency‐based and focus on relationship‐building (active listening and sensitive one‐to‐one conversations), parent engagement (communicating beyond grades and managing difficult discussions), classroom climate management (setting inclusive norms, bullying prevention, and conflict resolution), and mental health literacy and crisis response (recognizing warning signs, de‐escalation, and referral pathways). These competencies are behavioral and context‐specific, requiring practice‐based training rather than traditional lectures. Effective formats include role‐playing difficult scenarios, case discussions facilitated by experienced teachers, and feedback on video recordings of actual classroom or parent interactions. Second, schools can accelerate learning‐by‐doing through structured mentoring and supervised early leadership opportunities. For example, novice head teachers can be paired with high‐performing experienced mentors for classroom observation, co‐leading parent meetings and student interventions with real‐time coaching, and guided practice with systematic debriefing.

Funding

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting Information S1

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.